Errors in identification using natural markings: rates, sources, and effects on capture–recapture estimates of abundance

Abstract

The results of a double-marking experiment using natural markings and microsatellite genetic markers to identify humpback whales (Megaptera novaeangliae) confirm that natural markings are a reliable means of identifying individuals on a large scale. Of 1410 instances of double tagging, there were 414 resightings. No false positive and 14 false negative errors were identified. The rate of error increased with decreasing photographic quality; no errors were observed among photographs of the highest quality rating, whereas an error rate of 0.125 was identified in sightings for which only part of the area used for identification was visible. There was also a weaker relationship between error rate and the distinctiveness of markings, which may result from non-independence in coding for image quality and distinctiveness. A correction is developed for the Petersen two-sample abundance estimator to account for false negative errors in identification, and a parametric bootstrap procedure for estimation of variance is also developed. In application to abundance estimates from the North Atlantic, the correction reduces the bias in estimates made using poorer quality photographs to a negligible level while maintaining comparable precision.Résumé : Une expérience de marquage double à l'aide de marques corporelles et de marqueurs génétiques microsatellites confirme que l'utilisation des marques corporelles est une méthode fiable d'identification à grande échelle des individus du Rorqual à bosse (Megaptera novaeangliae). Des 1410 événements de double marquage, 414 impliquaient des animaux vus antérieurement. Aucun résultat faussement positif ou faussement négatif n'a été décelé. Le taux d'erreur augmentait avec la mauvaise qualité des photographies; aucune erreur ne s'est produite à l'utilisation des photographies de haute qualité et le taux d'erreur était de 0,125 lorsque seulement une partie de la surface de l'animal utilisée pour l'identification était visible. Il y avait aussi une relation plus faible entre le taux d'erreur et le caractère distinctif des marques corporelles, qui peut provenir du manque d'indépendance entre le codage de la qualité de l'image et le caractère distinctif de l'animal. On trouvera ici une correction de l'estimateur d'abondance à deux échantillons de Petersen pour tenir compte des résultats faussement négatifs dans l'identification, ainsi qu'une procédure bootstrap paramétrique pour estimer la variance. Appliquée à des estimations d'abondance provenant de l'Atlantique nord, la correction réduit à un niveau négligeable l'erreur due à l'utilisation de photographies de qualité inférieure dans les estimés, tout en maintenant un niveau de précision similaire.[Traduit par la Rédaction] 1870 Stevick et al.

Misidentification error associated with classifications of white‐tailed deer images

“…Consequently, the overall reliability of information from camera surveys will likely vary according to the experience level of those involved in processing images in situations where animals do not exhibit highly discernable differences in appearances. Although the observed effect size of the significant experience‐related factors was relatively small, any significant source of error is important because low rates of misidentification are known to bias estimates (Stevick et al , Morrison et al ). There currently are inconsistencies and analytical problems associated with methods used to collect and analyze camera‐trap data, and researchers have made calls for the refinement of methodologies to address these issues (Foster and Harmsen , Burton et al ).…”

Population dynamics of oceanic manta rays (Mobula birostris) in the Raja Ampat Archipelago, West Papua, Indonesia, and the impacts of the El Niño–Southern Oscillation on their movement ecology

“…A sighting was defined as a capture record on a specific site on a specific day; a recapture was defined as a sighting of a previously identified individual on a different day, or on a different site at least one‐hour post‐initial sighting. Similar to Stevick, Palsbøll, Smith, Bravington, and Hammond (), very low‐quality images were discarded to reduce identification errors. Secondary verification of IDs was performed manually by an outside expert and a third time using the Manta Trust’s “IDTheManta” matching software by trained research staff (The Manta Trust, ).…”

Abundance Estimates of Guiana Dolphins (Sotalia guianensis; Van Bénéden, 1864) Inhabiting an Estuarine System in Southeastern Brazil

“…Since then, mark-recapture methods have lected in summer (January 2015 and February been used to investigate ecological aspects of 2016) and winter (July and August 2016). Summer different populations (see Simão et al, 2000; and winter were chosen to contrast the extreme Santos et al, 2001;Azevedo et al, 2004; differences in the environment between these two & Bazzalo, 2004;Rossi-Santos et al, 2007;Hardt seasons, which could affect the way dolphins use et al, 2010; Batista et al, 2014). However, a the local estuary (Santos & Rosso, 2007).…”

“…The correct identification of marked inditograph all individuals in a group; and (5) capviduals is one of the most important assumptions to ture heterogeneity was incorporated in the fitted be met in mark-recapture studies as identification models. Researchers involved in fieldwork and errors lead to negatively (error type I) and posiphoto analyses did not change during data coltively (error type II) biased estimates of abundance lection, following strict criteria for including new (Gunnlausson & Sigurjónsson, 1990;Stevick et al, individuals in the catalog.…”